The use of compact transistors in integrated circuits requires that the dielectric layer present between the channel and the gate electrode is thin. To realize continuous shrinkage, it is essential to replace silicon dioxide (SiO2) as the dielectric material.
This is because at the thickness needed for the transistor design, the leakage current would become undesirable. In view of this reason, major advancements have been made towards materials exhibiting a high dielectric constant, also known as high-k dielectrics.
Such materials are typically built on hafnium. The transition from SiO2 brings forth certain analytical requirements. Besides the layer thickness, the following parameters become imperative:
- The distribution of elements in the layer
- Quantity of the active material in the layer
- The uniformity of these parameters across the wafer
- Thickness of any intermediate layer (SiO2 or metal silicate)
- Chemical state of the elements in the intermediate layer
The width of these dielectric layers is below the sampling depth of XPS. Hence, it is possible to study the entire dielectric layer at near normal electron emission angles sans the need for removing any material.
XPS gives data regarding the chemistry of the layers and their interfaces, whereas ARXPS gives further data rereading the thickness of layer and the dispersion of materials inside the layer. This article presents the data available from ARXPS and XPS with regard to high-k dielectric layers.
Relative Depth Plot
Relative depth plot is a simple technique that can be used for investigating samples containing more than a single layer on the substrate. This technique creates a chart on which the arrangement of the elements with depth is plainly displayed. Figure 1 illustrates a case of a relative depth plot from a combined layer of hafnium oxide (HfO2) and aluminum oxide on SiO2 on silicon.
Carbon can be observed close to the surface, while silicon is deep within the structure, as anticipated. The aluminum and hafnium occur at the same approximate position, and oxygen exists in two chemical states, as indicated by the two O 1s binding energies in the XPS spectrum; the high binding energy is related to hafnium and the low binding energy is related to silicon and aluminum.
Figure 1. Relative depth plot from a mixed HfO2, Al2O3 layer on SiO2 on Si
A feature of the Avantage data system is a multi-layer thickness calculator, which enables concurrent measurement of thickness for three thin layers on a substrate. The results of utilizing this method to determine the thickness of layers on a silicon substrate are shown in Figure 2.
Figure 2. Thickness of carbon, aluminum oxide and SiO2 layers on a silicon substrate, as a function of the number of ALD cycles. Thickness was determined using the multi- layer thickness calculator in Avantage.
Prior to growing aluminum oxide using the atomic layer deposition (ALD) technique, all samples had a thermal oxide layer grown on them. Samples also had an adventitious carbon layer. The plot illustrates that the thickness of Al2O3 layer rises with increasing number of ALD cycles, whereas the thickness of SiO2 layer remains constant.
In the presence of Al2O3, the thickness of the adventitious carbon is higher than it is on the SiO2 but does not depend on the thickness of the Al2O3 layer. In Figure 2, the shape of the Al2O3 line has been validated using other methods, such as ERD and TEM.
One way to determine the uniformity of the high-k layer is through mapping or a line scan measurement. Figure 3 shows a thickness line scan across the entire diameter of the wafer. The line scan distinctly shows that the thickness reduces by roughly 0.7nm.
Figure 4 shows Hf 4f and Al 2p maps obtained from a 200mm wafer on which a combined layer of HfO2 and Al2O3 was grown. The growth conditions were intentionally selected to promote an uneven growth. The maps apparently show the unevenness of the growth.
Figure 3. Thickness line scan across the diameter of the 200mm wafer showing the variation of the thickness of the mixed Al2O3 and HfO2 layer and the thickness of the SiO2 interfacial layer.
Figure 4. XPS maps of Al 2p (upper) and Hf 4f (lower) from a 200mm wafer.
Chemical State Maps
Oxygen related to hafnium is in an entirely different chemical state when compared to oxygen related to silicon and aluminum. A wafer map of oxygen in each of these two chemical states is shown in Figure 5.
Figure 5. XPS maps of O 1s. The lower map is oxygen in a state with a low binding energy (usually associated with hafnium). The upper map is oxygen in a state with a high binding energy (usually associated with aluminum and silicon).
Interfacial Layer Chemistry
When silicon is deposited with high-k dielectrics, an interfacial layer will exist between the silicon substrate and the layer. The nature of this layer will have certain effect on the layer’s electrical properties. The interfacial layers often include silicon in an oxidized state, a silicate or an oxide. The following factors may determine whether silicate is formed or not:
- The deposition temperature
- The method of preparation of the layer
- The treatment of the wafers after the deposition of the layer
- The nature of the surface on which the high-k layer is deposited
The binding energy of the Si 2p3/2 peak from hafnium silicate is 102.7eV, while the binding energy of the same peak from SiO2 is 103.6eV.
Figure 6. Comparison of Si 2p spectra from silicon dioxide (nominally 1.0 nm), with HfO2 grown by ALD and MOCVD on the same material.
Figure 6 illustrates the comparison between the Si 2p spectra from 1nm thick SiO2, 1nm SiO2 with 1nm thick HfO2 grown by ALD, and 1nm SiO2 with approximately 1nm thick HfO2 grown by metal organic chemical vapor deposition (MOCVD).
The spectra had a Shirley background deducted and normalized to the elemental peak position. From these spectra, it is evident that the oxidized silicon remained as SiO2 with little silicate formation because the binding energy continues to remain at 103.6eV and no substantiation exist for a low binding energy shoulder on the peak.
An analogous conclusion can be drawn for HfO2 grown on 0.5nm SiO2. The results, however, are different when the HfO2 is grown on thinner layers of SiO2. Figure 7 illustrates a comparison of Si 2p spectra from thin SiO2, a hydrofluoric acid etched silicon surface with HfO2 grown by ALD, and a thin oxide with HfO2 grown by ALD.
Figure 7. Comparison of Si 2p spectra from thin silicon dioxide (nominally 0.3 nm), with HfO2 grown by ALD and an HF-last surface with HfO2 grown by ALD.
Hydrofluoric acid etching aids in removing SiO2 from a silicon wafer, and a wafer treated in this manner will be called an HF-last surface. The thin thermal SiO2 can be clearly seen as an oxide san any silicate formation.
Upon growing HfO2 on 0.3nm SiO2, a distinct shift of the peak maximum can be seen and a widening of the peak indicates the presence of some amount of silicate. On the HF-last surface, the peak is moved more further to lower binding energies, indicating the presence of HfO2.
Figure 8 demonstrates the effect of growing a 4nm HfO2 layer by MOCVD on a substrate at 300°C temperature during growth, and the impact of annealing at 700°C temperature after growth. Based on these spectra, it looks as if there is a formation of silicate following the growth and that the amount of silicate increases following annealing.
A large amount of the oxidized silicon, however, remains as SiO2. Similar effects were seen during the growth of identical layers at 485°C temperature followed by annealing at 700°C.
If the ALD technique is used for growing the HfO2 layer, rather than MOCVD, the chemistry is entirely different (Figure 9). In this example, there is little formation of silicate after the actual growth. The proportion of silicate increases after the annealing step, but does not become as significant as that when the layer is prepared through the MOCVD method (Figure 8).
Figure 8. Comparison of Si 2p spectra from thin silicon dioxide (nominally 1.0nm), with 4nm HfO2 grown by MOCVD on a substrate at 300 °C and following annealing at 700 °C.
Figure 9. Comparison of Si 2p spectra from thin silicon dioxide (nominally 1.0nm), with HfO2 grown by ALD on a 1nm RTO and following annealing at 700 °C.
By means of techniques that involve the highest entropy, construct concentration depth profiles can be built from ARXPS data. An example of a depth profile via a mixed HfO2/Al2O3 on SiO2 on silicon sample is shown in Figure 10, and a reconstructed profile via HfO2 on SiO2 on silicon is illustrated in Figure 11.
Figure 10. Profile through mixed Al2O3/HfO2 on SiO2 on SI
Figure 11. Example of a depth profile through a sample HfO2 on SiO2 on Si. The profile was constructed from ARXPS data.
This profile demonstrates the presence of oxygen in two chemical states and also the presence of carbon at the surface and the estimated elements. The high binding energy O 1s peak is likely to be acquired from SiO2 and the oxygen present in adsorbed materials, whereas the low binding energy O 1s peak is believed to be obtained from the HfO2.
Comparison with Sputtering
Other methods available for studying these films involve sputter profiling with an ion beam. However, such a method may promote changes in the chemical states of the layer components.
Figure 12 depicts the changes in the chemical state of Hf due to sputtering. The sample included a 2.6nm layer of HfO2 on SiO2 on silicon. The layer of HfO2 was sufficiently thin that metallic Hf can be observed in the spectrum, provided it is present at the beginning of the experiment.
The initial spectrum in the montage was obtained prior to sample sputtering, and the subsequent spectrum was obtained after the material was removed by sputtering for a short period of time with 500eV argon ions.
Additional sputtering was performed after the acquisition of individual spectrum until a large amount of the Hf was removed. The final spectrum obtained in this series was the spectrum labeled 'interface'. It is obvious from this series of spectra that on account of sputtering, the HfO2 is being changed into hafnium metal.
Figure 12. Montage of XPS spectra from Hf 4f region during a sputter profile experiment.
Thermo Scientific Theta Probe and Theta 300 instruments give critical data for advanced gate dielectrics such as thickness of layer, thickness of the intermediate layer, uniformity of the layers, chemical states of the layer and the intermediate layer, and distribution of the material indie the layer.
As a non-destructive technique, ARXPS presents the use of sputtering with an ion beam. Sputtering not only changes the layer’s composition, but also causes atomic mixing that could lead to inaccurate data.
This information has been sourced, reviewed and adapted from materials provided by Thermo Fisher Scientific – Materials & Structural Analysis.
For more information on this source, please visit Thermo Fisher Scientific – Materials & Structural Analysis.